📄 Die Henne oder das Ei? azzo oder -zo? Weitere Spekulationen zu einem pronominalen Problem des Harari

Summary

Ewald Wagner revisits an open question in Harari historical grammar. Which came first: the 3rd person singular masculine possessive suffix -zo, or the independent pronoun azzo? One camp (Cohen, Cerulli, Leslau, Beniam Mitiku Cherinet) derives azzo from -zo. The other (Appleyard, Huehnergard and Pat-El) treats -zo as a worn-down azzo. Both forms are already fixed in the oldest Harari texts, written in Arabic script and probably dating to the late 16th century. The two stages of the language therefore cannot settle the matter directly.

Wagner instead reconstructs a "Proto-Harari" pronoun system and compares it with modern Harari and the related East Gurage language Silṭe. He brings in adpositions, word order, relative clauses and the areal context. He favours -zo as the starting point and lays out a five-stage path leading to azzo. A final section sketches, as less plausible, how azzo could have arisen without a prior -zo. The study shows how scattered textual traces can support a systemic reconstruction.

Key points

• Harari belongs to South Ethio-Semitic in Hetzron's classification. Its closest relatives are Silṭe, Wolane and Zway. It is attested as old Harari (ʿaǧamī, Arabic script) and modern Harari (fidäl script).
• Giorgio Banti found a manuscript of Kitāb al-Farāʾiḍ bearing an owner's note dated 1080 AH (1669). This supports dating the old literature to the second half of the 16th century.
• Reconstructed paradigms put Proto-Harari forms beside modern Harari and Silṭe ones. Examples are ḥuwa / azzo / uha for the pronoun and ()-o / -zo / -ka for the suffix.
• Two observations point to a native (non-Arabic) ḥuwa. Old texts often spell it ḤWʾ, unlikely for an Arabic loan. Burton (1856) records huwa together with a plural hiyyásh.
• Several -o endings in old Harari relative clauses can be read as traces of an older possessive suffix. These had been confusable with the vocative -o and with rhyme-filling -o.

Interesting ideas

• I1. Wagner keeps several readings of one ending open (vocative, rhyme filler, possessive) and decides between them by syntactic position, such as resumptive slots in relative clauses. — For Mova (tokens): the morphology layer could keep competing analyses of ambiguous affixes and let the parse context pick one, logging the deciding cue for explainability.
• I2. Spelling choice serves as evidence of origin: an unusual orthography argues that a word is native rather than borrowed. — For Mova (tokens): spelling variants and non-standard orthography could be kept as tagger features that flag likely loanwords or foreign tokens, rather than being normalised away.
• I3. The two rival hypotheses are compared by how much each still leaves to explain: one must account for both forms, the other only for azzo. — For Mova (frames): competing coreference or role assignments could be ranked by how many unsupported extra assumptions each requires, with that count shown in the explanation.
• I4. A single-morpheme question is tested for consistency against the whole pronoun system and related constructions, not in isolation. A counterexample in his own earlier data prompts the author to revise his old analysis. — For Mova (learn): self-induced repair rules could be checked against whole paradigms and held-out counterexamples before adoption, acting as negative controls.